A digital supply chain procurement transaction platform and method

Abstract

The application discloses a kind of digital supply chain procurement transaction platform and method, it is related to computer technology field, the platform includes NLP service cluster, structured order generation module, multidimensional intelligent matching engine, dynamic credit evaluation engine, unified identity authentication and authority management service module, ability portrait database, block chain distributed account book, RESTful API interface and integrity alliance cross early warning module;The method comprises: building the ternary collaborative architecture of purchaser, supplier and manufacturer;Parsing procurement demand and generating structured order;Parsing procurement demand and generating structured order;Generate dynamic credit evaluation report;Perform end-to-end closed-loop transaction management.The present application aims to solve the problem of single collaborative subject, low matching precision and credit evaluation lag of traditional procurement platform, improve supply and demand response timeliness, matching accuracy and transaction credibility, reduce the rate of dispute and supply chain interruption risk, support cross-industry efficient reuse.

Description

Technical Field

[0001] This invention relates to the field of computer technology, and in particular to a digital supply chain procurement and transaction platform and method. Background Technology

[0002] With the deepening development of the digital economy, supply chain procurement transactions are accelerating their transformation towards digitalization, platformization, and intelligence. Building efficient and transparent collaborative platforms has become a key means to improve resource allocation efficiency and supply chain resilience. Currently, procurement activities have evolved into a complex process encompassing demand forecasting, supplier selection, and contract fulfillment coordination, relying on accurate information exchange and reliable transaction guarantees. However, existing technological solutions still have significant shortcomings in areas such as multi-entity collaborative mechanisms, intelligent matching accuracy, credit assessment systems, and closed-loop management throughout the entire process, making it difficult to meet the dual demands for efficiency and reliability in procurement transactions under complex industrial scenarios.

[0003] Existing digital procurement platforms mainly fall into two categories: one is a comprehensive platform centered on process control. While it achieves transparency in the procurement process and real-time data transmission, it lacks structured support for deep collaboration among buyers, suppliers, and manufacturers. Furthermore, its matching logic does not incorporate multi-dimensional optimization algorithms based on capacity, quality, and historical performance records, leading to crude supply-demand matching and frequent resource mismatches. The other category is customized platforms for vertical industries. Although they have built automatic matching models, their highly vertical design sacrifices platform versatility, and the matching decision-making criteria are unclear. They fail to link inventory, capacity, and logistics resources on the production side, remaining only at the order matching level. More importantly, neither type of solution establishes a dynamic credit assessment and feedback mechanism covering the entire transaction cycle, making it difficult to support long-term, stable, and traceable supply chain partnerships.

[0004] Therefore, there is an urgent need for a new type of digital supply chain procurement and transaction platform and method that can integrate a three-entity collaborative architecture, multi-dimensional intelligent matching algorithms and a whole-process trusted execution mechanism to achieve efficient, transparent and closed-loop transactions from end to end. Summary of the Invention

[0005] The purpose of this invention is to provide a digital supply chain procurement transaction platform and method to solve the problems of single collaborative entities, low matching accuracy, and lagging credit assessment in existing traditional procurement platforms.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] On the one hand, a digital supply chain procurement and transaction platform includes:

[0008] The NLP service cluster receives unstructured procurement requests submitted by buyers and extracts material codes, specifications, quantities, delivery time windows, and key fields of quality requirements through pre-trained domain-specific language models.

[0009] The structured order generation module connects to the NLP service cluster and combines industry standard knowledge base to perform semantic verification and standardization transformation on the extracted fields, generating structured purchase orders that conform to the platform specifications.

[0010] The multi-dimensional intelligent matching engine connects the structured order generation module with the capability profile database. Based on historical fulfillment rate, real-time capacity utilization rate, raw material inventory level, quality pass rate, geographical location and logistics timeliness indicators, it uses a weighted linear combination and constraint satisfaction algorithm to calculate the matching score and screen candidate suppliers.

[0011] The dynamic credit assessment engine continuously collects order fulfillment data, quality inspection results, payment records, and third-party mutual evaluation information throughout the entire transaction cycle. It constructs an exponential smoothing scoring model through time decay factors and behavioral weight coefficients to update the credit rating of each participant in real time.

[0012] The unified identity authentication and access control service module, based on the OAuth2.0 protocol and RBAC model, provides single sign-on and fine-grained access control for capability profile fields for three types of entities: purchasers, suppliers, and manufacturers.

[0013] The capability profile database connects to the unified identity authentication and access control service module, storing dynamic capability profile data of buyers, suppliers, and manufacturers.

[0014] The blockchain distributed ledger connects the operation log interfaces of various modules, storing operation logs, contract signing, performance status and credit changes in an immutable manner;

[0015] The RESTful API interface provides standardized data interaction and communication support for each module.

[0016] The digital supply chain procurement and transaction platform also integrates a capacity scheduling optimization engine. When multiple concurrent purchase orders are received, it generates a globally optimal production scheduling scheme based on the manufacturer's equipment availability, process route, changeover time, and energy consumption constraints, using a mixed integer linear programming algorithm, with a scheduling conflict rate lower than a preset threshold.

[0017] The digital supply chain procurement and transaction platform supports cross-industry deployment and can be adapted to multiple industries by configuring an industry template library. The industry templates include material classification systems, quality standard sets, delivery cycle benchmarks, and credit evaluation rules, and the template switching response time is within a specific range.

[0018] The digital supply chain procurement and transaction platform has a built-in integrity alliance cross-early warning module. When a supplier's credit score is lower than the preset credit threshold multiple times in a row, or the capacity utilization rate exceeds a certain upper limit for a predetermined period of time, the system will automatically push alternative supply suggestions to the buyer and freeze the automatic confirmation function of high-risk transactions until they are manually reviewed and approved.

[0019] A digital supply chain procurement transaction method includes the following specific steps:

[0020] Step S1: Construct a three-element collaborative architecture: Through a unified identity authentication and access control service module, integrate the digital identities and capability profiles of three types of entities—buyers, suppliers, and manufacturers—into the platform, store them in the platform's capability profile database, establish a unified data interaction interface and access control mechanism, and realize structured collaboration among the three parties in the stages of demand release, capacity response, order confirmation, production scheduling, logistics delivery, and credit feedback.

[0021] Step S2: Parse procurement requirements and generate structured orders: Receive unstructured procurement requests submitted by buyers, extract material codes, specifications, quantities, delivery time windows, and key fields of quality requirements through pre-trained domain-specific language models deployed in the NLP service cluster, and perform semantic verification and standardization conversion in conjunction with industry standard knowledge base to generate structured procurement orders that conform to platform specifications.

[0022] Step S3 performs multi-dimensional intelligent matching: Based on the structured purchase order, the supplier and manufacturer capability profile data stored in the capability profile database are called up. The quantitative indicators of multiple dimensions such as historical fulfillment rate, real-time capacity utilization rate, raw material inventory level, quality pass rate, geographical location and logistics timeliness are comprehensively evaluated. The matching score is calculated by using a weighted linear combination and constraint satisfaction algorithm, and the candidate suppliers with the highest matching degree are selected.

[0023] Step S4 generates a dynamic credit assessment report: Throughout the entire transaction cycle, the dynamic credit assessment engine continuously collects order fulfillment data, quality inspection results, payment records, and tripartite mutual evaluation information. An exponential smoothing scoring model is constructed using time decay factors and behavioral weight coefficients to update the credit rating of each participant in real time. This rating serves as the core basis for subsequent matching and risk control. The integrity alliance cross-early warning module pushes alternative supply suggestions to the buyer when the credit score continuously falls below a preset threshold.

[0024] Step S5 executes end-to-end closed-loop transaction management: After successful matching, an electronic contract is automatically generated and a production scheduling instruction is triggered. The warehousing and logistics system is linked to allocate transportation resources, and a third-party payment gateway is integrated to complete phased fund settlement. After delivery, an automatic quality inspection and verification module is started to carry out quality inspection and credit feedback processes, forming a complete closed loop from demand release to credit archiving.

[0025] In step S1, the ternary collaborative architecture adopts blockchain-based distributed ledger technology. All entities' operation logs, contract signings, performance status, and credit changes are stored on the blockchain in an immutable manner to ensure the traceability and non-repudiation of the transaction process.

[0026] In step S2, the pre-trained domain-specific language model deployed in the NLP service cluster has a training corpus covering a large number of historical procurement documents. It supports context-aware quantitative conversion of ambiguous expressions such as "deliver as soon as possible" and "high-quality materials," and the conversion accuracy reaches the predetermined accuracy requirements.

[0027] The weighted linear combination formula used in the multi-dimensional intelligent matching in step S3 is as follows:

[0028]

[0029] Where S is the matching score, R is the historical fulfillment rate, C is the current capacity utilization rate, I is the raw material inventory adequacy, Q is the recent quality pass rate, L is the logistics distance, T is the estimated delivery time, P is the price competitiveness, and H is the historical cooperation frequency; each weight coefficient to The sum is 1, and the weighting is dynamically adjusted according to the category of goods purchased. The weighting configuration is automatically determined by the platform's intelligent optimization engine based on historical transaction data.

[0030] In step S4, the dynamic credit assessment engine uses an exponential smoothing recursive formula:

[0031]

[0032] in Let t be the credit score at time t. Score the behavior at time t. The time decay factor has a value within a preset range; behavioral score. The score is calculated by weighting factors such as on-time performance rate, quality inspection pass rate, timely payment, and peer review star rating according to preset rules, and the score is within the predetermined range.

[0033] The end-to-end closed-loop transaction management in step S5 includes an automatic quality inspection and verification module. This module connects to the API of a third-party testing agency and automatically triggers a sampling and testing task after the goods arrive at the designated warehouse. The testing items and standards are automatically generated based on the quality requirements in the purchase order, and the test results are sent back to the platform in real time and used as input for updating the credit score.

[0034] Compared with the prior art, the beneficial technical effects of the present invention are as follows:

[0035] This invention breaks through the limitations of traditional platforms that only connect buyers and sellers by constructing a structured collaborative architecture that covers purchasers, suppliers, and manufacturers. It incorporates production-side resources into the transaction decision-making closed loop, enabling procurement demand to directly drive capacity scheduling and raw material preparation, significantly improving the timeliness of supply and demand response. The blockchain-based evidence storage mechanism ensures transparency and credibility throughout the entire operation, greatly improving the efficiency of dispute resolution.

[0036] This invention introduces multi-dimensional quantitative evaluation indicators and a dynamic weighting mechanism to overcome the shortcomings of existing platforms that rely on single price or static qualification screening, resulting in a significantly higher matching success rate than traditional solutions. The capacity scheduling driven by mixed integer linear programming further ensures the executability of the matching results, reducing the order cancellation rate to a low level.

[0037] This invention achieves real-time and objective credit assessment through full-cycle credit data collection and an index-smoothed scoring model, significantly reducing the incidence of credit fraud. The integrity alliance's cross-early warning module proactively identifies potential performance risks and intervenes in high-risk transactions in advance, effectively reducing the probability of supply chain disruptions.

[0038] The universal industry template library design of this invention enables the platform to be quickly deployed in different industry scenarios without the need to redevelop the core logic, significantly shortening the implementation cycle; the high-precision parsing capability of the natural language processing model for unstructured requirements significantly reduces the threshold for buyers to use the platform, and the platform adoption rate is significantly improved.

[0039] This invention automates the entire process from demand analysis to credit archiving, eliminating information silos and manual intervention points, and significantly reducing the processing time of a single transaction; the automatic quality inspection and payment linkage mechanism ensures the reliability of "payment upon delivery", greatly reducing the transaction dispute rate and significantly enhancing the overall resilience and operational efficiency of the supply chain. Attached Figure Description

[0040] Figure 1 This is a schematic diagram of the overall technical architecture of a digital supply chain procurement and transaction platform proposed in this invention;

[0041] Figure 2 This is a schematic diagram illustrating the specific steps of a digital supply chain procurement transaction method proposed in this invention. Detailed Implementation

[0042] The features and exemplary embodiments of various aspects of the present invention will now be described in detail. To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are merely intended to explain the present invention and not to limit the present invention. For those skilled in the art, the present invention can be practiced without some of these specific details. The following description of the embodiments is merely to provide a better understanding of the present invention by illustrating examples of the invention.

[0043] Reference Appendix Figure 1 , Figure 1 This paper presents a schematic diagram of the overall technical architecture of a digital supply chain procurement and transaction platform proposed in this invention. It clearly depicts the collaborative relationship among the three main entities—buyers, suppliers, and manufacturers—within the platform, as well as the data flow and interface interaction methods between each module. (See attached reference.) Figure 2 , Figure 2 This diagram illustrates the specific steps of a digital supply chain procurement and transaction method proposed in this invention. It details the complete process from demand release, order parsing, intelligent matching, credit assessment, closed-loop transaction to risk warning, and clarifies the logical order and data transmission path between each step.

[0044] Example 1

[0045] In the digital supply chain procurement transaction methodology, step S1 constructs a three-element collaborative architecture, deploying a unified identity authentication and access control service module at the platform's underlying layer. This module, based on the OAuth 2.0 protocol, enables single sign-on and fine-grained access control for all three parties. When purchasing agents, suppliers, and manufacturers first access the platform, they must complete a digital identity registration process, including uploading legal documents such as business licenses, legal representative information, bank accounts, and production qualification certificates. The identity is then cross-verified using the platform's built-in OCR recognition engine and the National Enterprise Credit Information Publicity System to ensure authenticity. Upon successful verification, the platform generates a unique digital identity identifier for each entity. This identifier is encrypted and stored using the SHA-256 hash algorithm and serves as the primary key for all subsequent data interactions and operational audits.

[0046] Building upon digital identities, the platform constructs dynamically updated capability profiles for each entity, stored in a capability profile database. The capability profile for buyers includes fields such as historical purchase category distribution, average order size, delivery time tolerance, and frequency of quality complaints; the capability profile for suppliers covers dimensions such as available material lists, warehouse capacity, logistics partner lists, and on-time delivery rate statistics; and the capability profile for manufacturers is refined to parameters such as equipment type lists, maximum daily production capacity, process route libraries, raw material safety stock thresholds, and energy consumption coefficients. Equipment types include CNC machine tools, injection molding machines, and stamping lines. All profile data is stored in structured JSON format in the capability profile database, a distributed NoSQL database, with field-level update triggers set to ensure data real-time performance.

[0047] The data interaction interface of the ternary collaborative architecture adopts the RES Tful API specification, defining a standardized request / response message body structure. For example, the request body of the procurement requirement publishing interface includes the following fields: demand_id (unique requirement ID), material_code (material code), spec_params (specification parameter dictionary), quantity (quantity), delivery_window_start (delivery window start timestamp), delivery_window_end (delivery window end timestamp), and quality_requirements (quality requirement structure). All interfaces enforce HTTPS transmission encryption and JWT token authentication, with the token validity period set to 3600 seconds, automatically expiring after timeout.

[0048] The unified identity authentication and access control service module's access control mechanism is based on an extension of the RBAC model. The platform pre-defines three basic roles: purchasing operator, supplier scheduler, and production planner, and supports enterprise administrators defining custom sub-roles. Each role is granted access to specific API endpoints (GET / POST / PUT / DELETE) and read / write permissions to corresponding fields in the capability profile database. For example, a supplier scheduler can only read the material code and delivery time window in their associated procurement requirements, but cannot see the buyer's company name; a production planner can write the real-time capacity utilization rate of their equipment, but cannot modify the historical fulfillment rate field. Access control policies are configured in YAML format and loaded into the API gateway, with policy matching and validation performed on each request.

[0049] Step S1 employs a three-element collaborative architecture based on blockchain distributed ledger technology. The platform integrates the Hyperledger Fabric consortium blockchain network, establishing three organizational nodes corresponding to the buyer consortium, supplier consortium, and manufacturer consortium, with each organization maintaining one or more peer nodes. All entities' operation logs, electronic contract signing records, order status change events, and credit score update logs are encapsulated as transaction proposals and submitted to the channel. After being sorted by Kafka consensus, they are written to the world state database. Key data undergoes serialization processing before being uploaded to the blockchain: operation logs are converted to Protocol Buffer format, and only the SHA-3 hash value of the contract PDF file is uploaded to the blockchain; the original files are stored in the IPFS distributed file system and associated with a CID. The immutability of on-chain data ensures that the complete operation trajectory can be traced at any time, and non-repudiation is achieved through a digital signature verification mechanism—each transaction is signed by the initiator's private key, and the recipient can use their public key to verify the signature's validity.

[0050] In the digital supply chain procurement transaction methodology, step S2 parses the procurement requirements and generates structured orders, with the platform deploying a dedicated NLP service cluster. This cluster contains multiple NLP microservices that receive unstructured procurement request texts submitted by buyers through a web interface or EDI interface. The text content may contain free-format descriptions such as "Need 5000 A3 model stainless steel screws, delivery as soon as possible, high-quality materials required" or "Order 20 tons of B7 series aluminum alloy profiles, conforming to GB / T5237.1-2017 standard, delivery before August 15th." The NLP microservices first perform text cleaning, removing HTML tags, special symbols, and irrelevant whitespace characters, followed by word segmentation and part-of-speech tagging, and using a BiLSTM-CRF model to identify named entities.

[0051] For material code extraction, the system uses an industry standard knowledge base for fuzzy matching. The knowledge base includes built-in international and domestic standard coding systems such as ISO, GB, and DIN, supporting synonym mapping, such as mapping "stainless steel screws" to level A3-70 under ISO4762. For specifications, the system parses the numerical value and unit combination, such as converting "20 tons" to quantity=20000, unit="kg"; and verifies the reasonableness of the parameters, such as ensuring that the profile length cannot be negative. The delivery time window parsing module handles fuzzy time expressions: "Deliver as soon as possible" sets a default window based on the historical average delivery cycle of the purchased goods category, such as 3 days for electronic components and 30 days for heavy machinery; assuming the year is 2023, "before August 15th" is converted to a UTC timestamp delivery_window_end=1692057600. Quality requirement fields are matched to predefined quality templates through a rule engine; for example, "high-quality materials" is converted to specific testing indicators based on industry templates, such as the tensile strength of stainless steel screws ≥700MPa and salt spray test ≥96 hours.

[0052] The semantic validation phase performs multi-layered validation: the first layer is syntax validation, ensuring required fields are not empty; the second layer is logical validation, checking the matching of quantity and unit, such as "5000 pieces" corresponding to discrete materials and "20 tons" corresponding to continuous materials; the third layer is business rule validation, verifying whether the delivery time window is later than the current time but not shorter than the industry benchmark delivery cycle. If validation fails, a structured error code is returned, such as ERR_MISSING_MATERIAL_CODE and ERR_INVALID_DELIVERY_WINDOW, which the buyer can use to correct the request.

[0053] The final generated structured purchase order uses a platform-predefined XML Schema, including a root element and its child elements. After the order is generated, a globally unique order ID is assigned, with the format PO-YYYYMMDD-NNNNN, and is persisted to the order master table in the relational database. At the same time, an event bus is triggered to notify the downstream matching engine.

[0054] In step S2, a pre-trained domain-specific language model is deployed in the NLP service cluster. Based on the BERT architecture, this model undergoes domain-adaptive fine-tuning using 100,000 historical procurement documents accumulated by the platform, including quotations, contracts, and acceptance reports, on top of general Chinese pre-trained weights. Fine-tuning tasks include material entity recognition, delivery time extraction, and quality requirement classification. The model is deployed on a GPU-accelerated server, with inference latency controlled within 200 milliseconds. For quantization conversion of ambiguous expressions, the model integrates an attention mechanism to capture contextual cues—for example, in the context of "deliver as soon as possible" appearing in an emergency repair scenario, the model outputs a shorter delivery window of 24 hours; if it appears in a regular replenishment scenario, it outputs a longer window of 7 days. The conversion accuracy reached 98.5% as evaluated on the internal test set, meeting the predetermined accuracy requirements.

[0055] In the digital supply chain procurement transaction methodology, step S3 performs multi-dimensional intelligent matching. The platform activates the matching scheduler to listen for newly generated structured purchase order events. The matching scheduler first loads real-time capability profile data for all registered suppliers and manufacturers from the capability profile database, filtering out candidates whose capability range does not cover the order's material code. The remaining candidates enter the quantitative evaluation stage, where the system collects indicators across the following eight dimensions:

[0056] R stands for Historical Fulfillment Rate: This is calculated as the percentage of orders completed by the candidate party in the past 12 months that were delivered on time and to the required quality. The data is sourced from platform transaction records, and the calculation formula is... ,in To ensure timely delivery of orders, This represents the total number of orders, with a value range of [0, 1].

[0057] C represents the current capacity utilization rate: This is calculated by the platform aggregating the weighted average utilization rates of each piece of equipment, based on the manufacturer's real-time reports of production line status via IoT gateways. The weights are set according to the criticality of each equipment in the process flow. For example, the weight of equipment in the injection molding stage is 0.6, and that in the assembly stage is 0.4. A C value exceeding 0.95 is considered high load.

[0058] I refers to raw material inventory adequacy: The candidate's ERP system synchronizes raw material inventory data via API. The platform compares the required raw material quantity for the order with the current inventory quantity and calculates... This refers to the minimum inventory satisfaction rate of all necessary raw materials, with a value of [0,1].

[0059] Q refers to the recent quality pass rate: the percentage of qualified products in the candidate's most recent 30 batches of quality inspection reports, with data sourced from the platform's quality inspection module or third-party testing agency APIs. .

[0060] L represents the logistics distance: Based on the GPS coordinates of the candidate's warehouse registered on the platform and the buyer's receiving address coordinates, the shortest driving distance is calculated using the Gaode Map API and then normalized. ,in To ensure the industry's maximum reasonable delivery distance .

[0061] T stands for Estimated Delivery Time: The candidate predicts the delivery time based on the current production schedule and the ETA (Estimated Delivery Amount) of the logistics partner, and the platform calculates the time. If T≤1, then ,otherwise ,make sure .

[0062] P stands for price competitiveness: the reciprocal of the ratio of the candidate's price to the industry average price. After being cut off Then linearly map to [0,1].

[0063] H represents historical cooperation frequency: the number of historical orders between the candidate party and the current buyer, logarithmically compressed. ,make sure .

[0064] After each indicator is normalized to the [0,1] interval using Min-Max, the matching score S is calculated by substituting it into the weighted linear combination formula:

[0065]

[0066] Among them, the weighting coefficient The sum is 1, with initial values ​​preset by industry templates and dynamically adjusted by the platform's intelligent optimization engine. Industry templates, such as those for the electronics industry, emphasize Q and T with higher weights; while those for the building materials industry emphasize P and L. The intelligent optimization engine employs an online learning mechanism, collecting actual performance results (timeliness or quality inspection pass rate) for each completed transaction. Weights are updated using gradient descent to maximize the matching success rate. The constraint satisfaction algorithm synchronously performs hard constraint filtering: if a candidate's credit score is below a threshold (e.g., Cr < 0.6), or their capacity utilization rate (C > 0.98) persists for more than 24 hours, they are directly excluded and not included in the S calculation.

[0067] The matching scheduler sorts the candidates in descending order of their S values ​​and selects the top 5 as a recommendation list to return to the buyer's interface. The buyer can view the detailed capability profiles and historical transaction records of each candidate, and after confirming the selection, the closed-loop transaction process in step S5 is triggered.

[0068] In the digital supply chain procurement transaction methodology, step S4 generates a dynamic credit assessment report. The platform deploys a dynamic credit assessment engine to continuously monitor the entire transaction lifecycle event flow. Event sources include: order status changes, quality inspection results, payment gateway callbacks, and third-party mutual evaluations. Order status changes include creation, confirmation, production in progress, shipment, and receipt; quality inspection results include pass / fail and detailed reports; payment gateway callbacks include payment success / failure and timestamps; and third-party mutual evaluations include 1-5 star ratings and textual reviews from the buyer to the supplier, the supplier to the manufacturer, and the manufacturer to the buyer.

[0069] The dynamic credit assessment engine maintains a credit profile for each entity, with the core being the dynamic credit score Cr(t). This score employs an exponential smoothing scoring model, and its recursive formula is as follows:

[0070]

[0071] in The time decay factor is 0.3, falling within the preset range of [0.2, 0.5], ensuring that recent behavior has a greater impact on the score. B(t) is the behavior score at time t, calculated by weighting four sub-scores: on-time performance score. Based on the difference between the actual delivery time and the promised delivery time, it is linearly mapped to [0,1]; quality inspection pass rate score. Percentage of qualified batches, [0,1]; Payment timeliness score The reciprocal countdown of the payment delay days, compressed to [0,1] using the Sigmoid function; peer review star rating. The weighted average of the three evaluations, with the weights set according to the roles of the evaluators, [0,1]. Ensure that the total score is within the range of [0,1].

[0072] The credit profile also includes credit rating labels, such as AAA, AA, A, B, and C, based on the Cr(t) threshold: Cr ≥ 0.9 is AAA, 0.8 ≤ Cr < 0.9 is AA, and so on. The credit report is generated regularly, specifically at 2 AM daily, and includes a scoring trend chart, detailed scores for each dimension, and a summary of recent negative events. This report serves as the core input for step S3 matching—when the matching engine queries candidate credit profiles, it prioritizes the latest Cr(t) value and triggers a risk warning when Cr(t) < 0.6.

[0073] The Integrity Alliance's cross-monitoring module operates synchronously, continuously monitoring the credit scores and capacity utilization rates of each entity. When a supplier's credit score falls below a preset threshold three times consecutively (e.g., Cr < 0.6), or its capacity utilization rate exceeds 95% for more than 24 hours, the system automatically pushes alternative supply suggestions to relevant buyers and freezes the automatic confirmation function for high-risk transactions of that supplier until it is manually reviewed and approved.

[0074] In the digital supply chain procurement transaction methodology, step S5 executes end-to-end closed-loop transaction management. After the buyer confirms the matching result, the platform immediately invokes the electronic contract generation service. This service fills in the legal clauses, liability for breach of contract, and intellectual property statement in a preset contract template based on order data, generates a PDF contract, and performs multi-party electronic signatures using CFCA digital certificates. The signing process uses a timestamp service to ensure legal validity, and the contract hash value is stored on the blockchain as evidence after signing.

[0075] The moment the contract takes effect, the platform triggers a production scheduling instruction to the selected manufacturer's MES (Manufacturing Execution System) interface. The instruction includes parameters such as the bill of materials, process route requirements, and delivery deadline. Simultaneously, the capacity scheduling optimization engine intervenes: when multiple concurrent orders target the same manufacturer, the engine solves for the globally optimal schedule based on a mixed-integer linear programming model. Decision variables include the start time and changeover sequence of each order on each piece of equipment; the objective function minimizes total delay and energy consumption; constraints cover equipment availability windows, process route sequence, changeover time matrix, and maximum shift hours. The solver uses the commercial Gurobi solver, and the scheduling conflict rate—the proportion of orders that cannot meet the delivery time window—is measured to be less than 0.5%.

[0076] The logistics collaboration module works synchronously: the platform calls the TMS (Transportation Management System) API to automatically allocate optimal transportation resources based on cargo volume and weight, delivery address, and carrier service capabilities. The generated waybill number is sent back to the platform and bound to the order. The payment process integrates third-party gateways such as Alipay, WeChat Pay, and UnionPay, supporting phased settlements—for example, a 30% prepayment triggered upon contract signing, a 60% payment triggered upon logistics receipt, and a 10% quality assurance deposit triggered upon quality inspection. Payment conditions at each stage are automatically monitored by smart contracts; a payment request is initiated when the conditions are met.

[0077] Upon delivery, the automated quality inspection and verification module is activated: when goods arrive at the designated warehouse, the WMS (Warehouse Management System) scans the waybill number, triggering a quality inspection task. Task parameters, including sampling ratio, testing items, and judgment criteria, are automatically generated strictly according to the quality requirements in the purchase order. The platform calls the open APIs of third-party testing organizations such as SGS and BV to push testing instructions and receive structured testing reports. The report includes fields such as measured values ​​of the tested items, whether they are qualified, and descriptions of non-conformities. This report is updated in real-time to the order status and serves as input events for the dynamic credit assessment engine, driving Cr(t) updates.

[0078] Thus, a closed loop is formed, encompassing the entire process from demand posting, matching, signing contracts, production, logistics, payment to quality inspection feedback. Data from all stages flows seamlessly within the platform without the need for manual intervention.

[0079] To illustrate the effectiveness of this invention, the following application example is constructed: A home appliance manufacturer (purchaser) submits a purchase request through the platform: "Urgently need 10,000 sets of air conditioner compressor brackets, made of ADC12 aluminum alloy, with powder coating, to be delivered before August 20th, and the quality must meet the QC / T518-2020 standard." The platform's NLP service cluster parses the request and generates a structured order:

[0080] material_code="COMP-2023-ALU-BRKT", quantity=10000, delivery_window_end=1692489600(2023-08-2000:00:00UTC), quality_requirements references industry template ID="AUTO-QC518".

[0081] The matching engine selected three candidate manufacturers from the capability profile database. Their capability profile data is as follows:

[0082] Manufacturer A:

[0083] ;

[0084] Manufacturer B:

[0085] ;

[0086] Manufacturer C:

[0087] .

[0088] Assuming the current weight configuration is w=[0.2,0.1,0.1,0.2,0.1,0.15,0.1,0.05] (emphasizing quality and fulfillment), the calculation yields...

[0089] The platform recommends Manufacturer C as the first choice. After the buyer confirms, the electronic contract is automatically generated and signed. The capacity scheduling engine checks Manufacturer C's production schedule and finds a gap in its die-casting line from August 10th to 15th, so production is arranged; the logistics module assigns SF Express for large items, with delivery expected on August 18th. After the goods arrive at the warehouse, the automatic quality inspection and verification module automatically triggers SGS testing, and the report confirms compliance with QC / T518 standards. The payment gateway releases the final payment, and the dynamic credit assessment engine updates Manufacturer C's Cr(t) value, improving its credit rating.

[0090] Example 2

[0091] In another implementation scenario, the platform is deployed in the cross-border electronic component procurement sector. The buyer, located in Germany, submits a procurement request in English.

[0092] "Need 50000 pcs of SMD resistors 0805 10 "±1%, RoHS compliant, delivery by Sep 10, 2023". The platform's NLP service cluster integrates a multilingual processing module. It first translates the text into Chinese using the Google Translate API, then executes the aforementioned parsing process to generate a structured order. During matching, the weight configuration automatically switches to the electronics industry template. Due to the extremely high requirements for quality and delivery time in the electronics industry, candidate suppliers include manufacturers in Shenzhen, China; Penang, Malaysia; and Hanoi, Vietnam. The logistics distance L has been normalized, where... Shenzhen manufacturers Penang scored 0.85, and Hanoi 0.88. Ultimately, the Shenzhen manufacturer won due to its highest overall score. Cross-border payments were completed through the PayPal gateway, and quality inspection was conducted by… Execution. Data is stored and verified on the blockchain throughout the entire process, meeting GDPR and cross-border data compliance requirements.

[0093] Example 3

[0094] In a bulk commodity procurement scenario, a steel mill needs to purchase 100,000 tons of iron ore. The procurement demand is expressed as follows: The NLP service cluster identifies "MT" as metric tons. Quality requirements are converted into specific chemical component thresholds. During matching, due to the involvement of maritime transport, the logistics timeliness (T) calculation includes port loading / unloading and shipping schedules, while L uses sea freight distance. Weighting is focused on P (price) and I (inventory) because commodity prices fluctuate significantly and a continuous supply of raw materials must be ensured. The platform connects to global port AIS data to calculate vessel ETA in real time. Credit assessment pays particular attention to historical large-value transaction performance records. Ultimately, the system matches the supplier with an Australian mining company; the contract uses letters of credit for payment, and quality inspection is performed by SGS at the destination port using wet tonnage testing. Capacity scheduling in this scenario is simplified to coordinating loading plans.

[0095] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A digital supply chain procurement and transaction platform, characterized in that, include: The NLP service cluster receives unstructured procurement requests submitted by buyers and extracts material codes, specifications, quantities, delivery time windows, and key fields of quality requirements through pre-trained domain-specific language models. The structured order generation module connects to the NLP service cluster and combines industry standard knowledge base to perform semantic verification and standardization transformation on the extracted fields, generating structured purchase orders that conform to the platform specifications. The multi-dimensional intelligent matching engine connects the structured order generation module with the capability profile database. Based on historical fulfillment rate, real-time capacity utilization rate, raw material inventory level, quality pass rate, geographical location and logistics timeliness indicators, it uses a weighted linear combination and constraint satisfaction algorithm to calculate the matching score and screen candidate suppliers. The dynamic credit assessment engine continuously collects order fulfillment data, quality inspection results, payment records, and third-party mutual evaluation information throughout the entire transaction cycle. It constructs an exponential smoothing scoring model through time decay factors and behavioral weight coefficients to update the credit rating of each participant in real time. The unified identity authentication and access control service module, based on the OAuth2.0 protocol and RBAC model, provides single sign-on and fine-grained access control for capability profile fields for three types of entities: purchasers, suppliers, and manufacturers. The capability profile database connects to the unified identity authentication and access control service module, storing dynamic capability profile data of buyers, suppliers, and manufacturers. The blockchain distributed ledger connects the operation log interfaces of various modules, storing operation logs, contract signing, performance status and credit changes in an immutable manner; The RESTful API interface provides standardized data interaction and communication support for each module.

2. The digital supply chain procurement and transaction platform according to claim 1, characterized in that, The digital supply chain procurement and transaction platform supports cross-industry deployment and can be adapted to multiple industries by configuring an industry template library. The industry templates include material classification systems, quality standard sets, delivery cycle benchmarks, and credit evaluation rules.

3. The digital supply chain procurement and transaction platform according to claim 1, characterized in that, The digital supply chain procurement and transaction platform has a built-in integrity alliance cross-early warning module. When a supplier's credit score falls below a preset credit threshold multiple times in a row, or when the capacity utilization rate exceeds a specific upper limit for a predetermined period of time, the system automatically pushes alternative supply suggestions to the buyer and freezes the automatic confirmation function of high-risk transactions.

4. The digital supply chain procurement and transaction platform according to claim 1, characterized in that, The digital supply chain procurement and transaction platform integrates a capacity scheduling optimization engine. When it receives multiple concurrent purchase orders, it generates a globally optimal production scheduling scheme based on the manufacturer's equipment availability, process route, changeover time, and energy consumption constraints using a mixed integer linear programming algorithm.

5. A digital supply chain procurement transaction method, applied to the digital supply chain procurement transaction platform described in any one of claims 1 to 4, characterized in that, include: Step S1: Through the unified identity authentication and access control service module, the digital identities and capability profiles of three types of entities—buyers, suppliers, and manufacturers—are integrated into the platform and stored in the platform's capability profile database. A unified data interaction interface and access control mechanism are established to achieve structured collaboration among the three parties in the stages of demand release, capacity response, order confirmation, production scheduling, logistics delivery, and credit feedback. Step S2: Receive unstructured procurement requests submitted by buyers, extract key fields through a dedicated language model in the NLP service cluster, and perform semantic verification and standardized transformation in conjunction with an industry standard knowledge base to generate structured procurement orders that conform to platform specifications. Step S3: Based on the structured purchase order, call the supplier and manufacturer capability profile data, comprehensively evaluate the quantitative indicators of multiple dimensions such as historical fulfillment rate, real-time capacity utilization rate, raw material inventory level, quality pass rate, geographical location and logistics timeliness, and use the weighted linear combination and constraint satisfaction algorithm to calculate the matching score and screen the candidate suppliers with the highest matching degree. Step S4: Collect order fulfillment data, quality inspection results, payment records and three-party mutual evaluation information through a dynamic credit assessment engine. Construct an exponential smoothing scoring model through time decay factor and behavioral weight coefficient, update the credit rating of each participant in real time, and use the credit rating as the core basis for subsequent matching and risk control. Step S5: After successful matching, an electronic contract is generated and a production scheduling instruction is triggered. The warehousing and logistics system is linked to allocate transportation resources, and a third-party payment gateway is integrated to complete phased fund settlement. After delivery, the automatic quality inspection and verification module is started to carry out quality inspection and credit feedback processes, forming a complete closed loop from demand release to credit archiving.

6. The digital supply chain procurement transaction method according to claim 5, characterized in that: Step S1 employs blockchain-based distributed ledger technology, where all entities' operation logs, contract signings, performance status, and credit changes are stored in an immutable manner to ensure traceability and non-repudiation of the transaction process. The unified identity authentication and access control service module is implemented based on a role-based access control model, with three basic roles: procurement operator, supplier scheduler, and production planner. Each role is granted operation permissions for specific interface endpoints and read / write permissions for capability profile fields. Policy matching verification is performed for each data interaction.

7. The digital supply chain procurement transaction method according to claim 5, characterized in that, The pre-trained domain-specific language model deployed in the NLP service cluster has a training corpus covering historical procurement documents. It supports context-aware quantization conversion of fuzzy expressions, and the conversion results meet the predetermined accuracy requirements.

8. The digital supply chain procurement transaction method according to claim 5, characterized in that, In the weighted linear combination formula used in step S3, the indicators for each dimension include historical fulfillment rate, current capacity utilization rate, raw material inventory adequacy, recent quality pass rate, logistics distance normalized value, expected delivery time score, price competitiveness normalized value, and historical cooperation frequency. Each indicator is normalized and then entered into the calculation. The sum of the weight coefficients is 1 and is dynamically adjusted according to the procurement category.

9. The digital supply chain procurement transaction method according to claim 5, characterized in that: In step S4, the dynamic credit assessment engine adopts an exponential smoothing recursive formula. The current credit score is obtained by weighting the current behavior score and the historical credit score. The behavior score is calculated based on the on-time performance rate, quality inspection pass rate, payment timeliness, and mutual rating star according to preset weights. The integrity alliance cross-early warning module pushes alternative supply suggestions to the purchaser when the credit score is continuously lower than the preset threshold.

10. The digital supply chain procurement transaction method according to claim 5, characterized in that, Step S5 includes an automatic quality inspection and verification module. After the goods arrive at the designated warehouse, the automatic quality inspection and verification module automatically triggers a sampling and inspection task. The inspection items and standards are automatically generated based on the quality requirements in the purchase order. The inspection results are transmitted back to the platform in real time and used as input for credit score updates.

Images